Exploring Applications of Second Generation Archived

Exploring Applications of Second Generation Archived Transit Data for Estimating

Performance Measures and Arterial Travel Speeds

Travis B. Glick

Department of Civil and Environmental Engineering

Portland State University

P.O. Box 751

Portland, OR 97207-0751

Phone: (503) 725-4282

Fax: (503) 725-5950

Email: [email protected]

Wei Feng

Performance Measurement

Chicago Transit Authority

Chicago, IL 60661

Phone: (312) 681-3486


Robert L. Bertini


California Polytechnic State University

1 Grand Avenue

San Luis Obispo, CA 93407-0353

Phone: (805) 756-1365

Fax: (805) 756-6330


Miguel A. Figliozzi


Portland State University

P.O. Box 751

Portland, OR 97207-0751

Phone: (503) 725-4282

Fax: (503) 725-5950


Forthcoming 2015 Transportation Research Record

ABSTRACT

Travel time and operating speed influence transit service attractiveness, operating cost, and system

efficiency. As part of their bus dispatch system (BDS), the Tri-County Metropolitan Transportation

District of Oregon (TriMet) has been archiving automatic vehicle location (AVL) and automatic

passenger count (APC) data for all bus trips at the stop level since 1997. In 2013, a new and higher

resolution bus AVL data collection system was implemented. This new system provides stop-level data as

well as 5-second resolution (5-SR) bus position data between stops. The objective of this paper is to

explore potential applications of the new data for assessing transit performance and for estimating

transportation system performance measures for urban arterials. Results suggest that the 5-SR data

provides high resolution time and position information which can be used to determine bus travel speeds

between bus stops, identify speed breakdowns, and estimate intersection signal/queuing delays.

Keywords: transit performance; archived data; automatic vehicle location; speed breakdown; urban

arterial

Glick, Feng, Bertini, and Figliozzi 2

INTRODUCTION

Analysis and modeling of bus travel time and operating speed are integral for the transportation planning

process (1), operations management, route planning and scheduling, and continued performance

measurement and evaluation (2). This study uses two sets of archived transit data provided by the Tri-

County Metropolitan Transportation District of Oregon (TriMet) via their Bus Dispatch System (BDS):

stop-level metrics including automatic passenger count (APC) data and new 5-second resolution data (5-

SR) automatic vehicle location (AVL) data between stops. The objective of this paper is to explore the

new 5-SR data for a particular bus route in Portland, OR and to analyze more detailed bus trip time and

speed information that are not visible using previously available, lower resolution AVL and APC data.

This paper examines finer resolution bus travel speed as a means to examine speed changes, queuing, and

delay at signalized intersections and other locations along route.

The use of buses as probes to estimate travel times has been studied in the past (e.g., 3,4). In

particular, TriMet buses have been used as probe vehicles (5) to evaluate arterial and transit performance

(6). However, these studies used TriMet stop-level data, which was all that was available at the time. To

estimate travel times and trajectories between stops, researchers had to use proxies, such as removing the

estimated stop time, which does not account for bus accelerations or decelerations, or estimating travel

time using the recorded/reported maximum speed in a segment between stops. Signal delay is also a key

source of variability for bus travel time (7) and is of interest to transit operators and researchers.

The route chosen for this study was a 14.8 mile segment that ran from NE Kelly & 5th to NW 6th

& Flanders in Portland, Oregon. Route 9 is shown in Figure 1. Due to some missing data, the first

westbound stop, Gresham Transit Center, was not included in the analysis. Route 9 included 71

westbound trips on 01 May 2013. A total of 22 trips traversed Powell Blvd. between 6:00 a.m. and 10:00

a.m.: fifteen trips began at NE Kelly & 5th, six trips began at SE Powell and 92nd, and one began at W

Powell & NW Birdsdale. All 22 buses ended their westbound trips at NW 6th and Flanders. Data for all

22 trips were extracted from both the AVL/APC and 5-SR systems.

(a) Route 9 schematic (source: TriMet Website)


(b) Route 9 (westbound) trip plotted onto Google Maps.

FIGURE 1 TriMet Route 9 overview.

DATA TriMet has implemented a Bus Dispatch System (BDS) as a part of its operation and monitoring control

system (8,9,10). The BDS post-processes and archives detailed stop-level data from buses during all trips.

This includes scheduled departure time, dwell time, actual arrival and departure times, and the number of

boarding and alighting passengers at every stop. The BDS also logs data for every stop in the system,

whether or not the bus stops to serve passengers. Error! Reference source not found.A contains a

sample list of archived BDS data. The calendar date, vehicle, badge, train, trip, and route number are all

listed for identification. In the far right column, a location ID number is listed. Each stop location has

GPS coordinates associated with it. These coordinates are the basis by which arrival, leave, and dwell

time are recorded. Times are not given as shown, but are presented in seconds past midnight, which is

then converted into a standard clock format.

Additionally, each location has a predefined 45 foot stop circle surrounding the stop. The “arrive

times” and “leave times” are recorded as the time the bus enters and exits the stop circle, respectively.

Dwell is recorded when the doors of the bus open. If dwell occurs, arrive time is recorded as the time the

doors open; arrive time plus the number of seconds of dwell gives the leave time. Figure 2 shows this

setup.

FIGURE 2 BDS Data – Stop circle definition.

When passenger activity occurs, the total number of boarding and alighting passengers is recorded in

two separate fields by APCs installed on front and rear door. APCs use infrared light to detect passenger

movement and are only activated if the doors open. The use of a lift to assist passengers with disabilities

is indicated in the lift field of the BDS data. Additionally, two distance measurements are recorded:

pattern distance and train mileage. Pattern distance is an estimate of the linear distance, in feet, from the

Dwell

Time Arrive time

Time Line

45 feet

Doors

Open Doors

Close

Arrive time if dwell occurs

Leave time

NE Kelly & 5th

NW 6th and Flanders

End Ross Island Bridge

Kelly & 5th

Begin Ross Island Bridge

Kelly & 5th SE 82nd Ave.

N


beginning of a route’s pattern to the vehicle’s current location. Note that scheduled trips for a single

vehicle are grouped together into blocks called "trains" for assignment. Train mileage is the cumulative

distance, in miles, from the start of the "train’s" recorded service (11).

TABLE 1 Sample TriMet Data

a) BDS Data Table Service

Date

Vehicle

Number Badge Train

Trip

Number

Stop

Time

Arrive

Time Dwell

Leave

Time Door Lift Ons Offs

Train

Mileage

Pattern

Dist. (ft.)

Location

ID

1-May-13 2260 1892 934 1140 6:54:30 6:54:58 0 6:54:58 2 0 3 2 26.63 0 3123

1-May-13 2260 1892 934 1140 6:54:53 6:55:32 0 6:55:32 0 0 0 0 26.74 568 7605

1-May-13 2260 1892 934 1140 6:55:31 6:56:23 10 6:56:23 0 0 0 0 26.92 1,496 13033 1-May-13 2260 1892 934 1140 6:56:19 6:57:02 0 6:57:02 0 0 0 0 27.12 2,671 12862

1-May-13 2260 1892 934 1140 6:56:36 6:57:12 0 6:57:12 0 0 0 0 27.19 3,097 9347

1-May-13 2260 1892 934 1140 6:56:59 6:57:28 0 6:57:28 0 0 0 0 27.30 3,668 4558 1-May-13 2260 1892 934 1140 6:57:38 6:57:56 16 6:58:12 2 0 1 0 27.49 4,606 12868

1-May-13 2260 1892 934 1140 6:58:09 6:58:31 0 6:58:31 0 0 0 0 27.63 5,377 12863

1-May-13 2260 1892 934 1140 6:58:39 6:58:49 0 6:58:49 0 0 0 0 27.77 6,122 4556 1-May-13 2260 1892 934 1140 6:59:10 6:59:20 15 6:59:20 0 0 0 0 27.91 6,870 4553

1-May-13 2260 1892 934 1140 6:59:49 6:59:44 0 6:59:44 0 0 0 0 28.10 7,835 12864

1-May-13 2260 1892 934 1140 7:00:31 7:00:09 0 7:00:09 0 0 0 0 28.29 8,871 4516

b) 5-Second Resolution Data with Calculated Values

Vehicle ID Opd Date Act Time *Time *Gap

Interval

GPS

Latitude

GPS

Longitude *∆ Distance

*Cum.

Distance

2205 1-May-13 23334 6:28:54 0:00:05 45.496690 -122.458707 0.0000 0.0000

2205 1-May-13 23339 6:28:59 0:00:05 45.496295 -122.459357 0.0417 0.0417

2205 1-May-13 23349 6:29:09 .0:00:10. 45.495643 -122.460055 0.0563 0.0980

2205 1-May-13 23419 6:30:19 .0:01:10. 45.495472 -122.460537 0.0262 0.1242

2205 1-May-13 23424 6:30:24 0:00:05 45.495165 -122.461137 0.0360 0.1601

2205 1-May-13 23429 6:30:29 0:00:05 45.494790 -122.461758 0.0397 0.1998

2205 1-May-13 23434 6:30:34 0:00:05 45.494400 -122.462355 0.0395 0.2394

2205 1-May-13 23439 6:30:39 0:00:05 45.494037 -122.463117 0.0446 0.2840

2205 1-May-13 23444 6:30:44 0:00:05 45.493823 -122.463692 0.0315 0.3155

2205 1-May-13 23464 6:31:04 .0:00:20. 45.493585 -122.464338 0.0354 0.3509

2205 1-May-13 23469 6:31:09 0:00:05 45.493212 -122.465003 0.0413 0.3921

The vehicle identifiers included in the BDS data are not available for the 5-second resolution data,

which records a timestamp and GPS location of each bus between stops every 5 seconds if the bus is

moving. Using the BDS data as a guide, 5-SR data can be extracted by comparing the times recorded on

each. 5-SR data does not start and stop at the beginning and end of a trip; by determining the start and end

times for any specific bus and day, that information can be used to define a complete trip of the 5-SR

data. Error! Reference source not found.B contains sample 5-SR data and additional calculated values.

Columns marked with an asterisk in the table were calculated. When the bus is not moving, data points do

not record every 5 seconds and Table 1B contains three gap intervals greater than 5 seconds. Hence, the

resolution of the 5-SR data is up to 5 seconds.

TRAJECTORY ANALYSIS

The trajectories of 22 westbound Route 9 buses are plotted in Figure 3 using a time-space diagram, where

the slope of each trajectory is the average speed of a bus. The figure shows cumulative distances versus

time for all 22 trips between 6:00 a.m. and 10:00 a.m. for Route 9 created using BDS data. The first

number in ####_#### is the bus number and the second number is the trip number.


FIGURE 3 Westbound Route 9 time space diagram

Figure 3 shows how the buses’ speeds vary over time (scheduled stop times are not shown for

clarity). This westbound route carries buses past 84 scheduled stops. On average, the buses stopped at 46

stops but as few as 33 and as many as 54 were serviced by any one bus. Similar trends can be seen across

all bus routes. For example, a sudden decrease in average speed is observed just before mile 4 when the

buses reach SE 162nd on Powell Blvd; further, a constant average speed is observed at mile 12 when

buses begin to travel across the Ross Island Bridge.

Previous research has examined methods for analyzing bus trip time and for producing transit

performance measures using archived stop level AVL/APC data for Route 14 in Portland (12,13).

Building on this and other previous research in the literature, this paper aims to test and modify/improve

the previous methods that relied on the stop level data using the higher resolution bus AVL data now

available. To ensure this comparison is justified, bus 2231/trip 1170 was analyzed in depth using both

data sets. A time-space diagram (Figure 4A) and an oblique curve (Figure 4B) were created. This

comparison highlights the small amount of deviation between the two data sets (14). This result was

constant across different buses and trips. Moving forward it can be assumed that the two data sets are

close enough to compare directly. Additionally, the similarities between the trajectory analyses implies

little benefit to using 5-SR data over stop level data when conducting a trajectory analysis. The previous

body of research (12,13) are sufficient to compare differences between scheduled and actual arrive times

at stop locations, which are the only locations where scheduled times are available.

a) Time-space diagram created using BDS and 5-SR data

22

05

_10

90

22

05

_12

60

22

31

_11

70

22

60

_11

40

22

60

_12

50

22

62

_10

70

22

62

_12

40

22

80

_10

60

22

80

_12

00

25

33

_10

50

25

44

_12

20

26

16

_11

20

27

02

_11

60

28

36

_12

70

30

50

_11

90

22

34

_11

10

22

90

_11

30

25

33

_11

80

25

44

_10

40

25

46

_10

80

25

46

_12

10

26

04

_11

50

0

3

6

9

12

15

Cu

mu

lati

ve D

ista

nce

(m

iles)

0

3

6

9

12

15

Cu

mu

lati

ve D

ista

nce

(m

iles)

5-SR Data

BDS Data

ScheduledTrajectory


b) Oblique Curve

FIGURE 4 Bus 2231 Trip 1170

CUMULATIVE ANALYSIS

Global Positioning System (GPS) coordinates can be used to calculate distances between two points.

Since GPS coordinates were recorded where −180° < 𝑙𝑜𝑛𝑔𝑖𝑡𝑢𝑑𝑒 < 180° and −90° < 𝑙𝑎𝑡𝑖𝑡𝑢𝑑𝑒 < 90°,

was used to calculate distances between two points. These differences were then added together for a

cumulative distance value. 3959 miles was used as it is the average radius of the Earth.

cos−1 (sin (𝑙𝑎𝑡1°∙𝜋

180°) ∙ sin (

𝑙𝑎𝑡2°∙𝜋

180°) + cos (


180°) ∙ cos (


180°) ∙ cos (

𝑙𝑜𝑛𝑔2°∙𝜋

180°−

𝑙𝑜𝑛𝑔1°∙𝜋

180°)) ∗ 3959 𝑚𝑖𝑙𝑒𝑠 (1)

When attempting to compare the AVL/APC data to 5-SR positioning data, only the bus number and

timestamps can be used to cross reference the data. Once equivalent timestamps have been established,

cumulative distance could be calculated for both data sets. The stop level AVL/APC data allows for

distance to be calculated three different ways. Train mileage, pattern distance, and stop location GPS data

result in average cumulative distances of 14.614 miles, 14.814 miles, and 14.487 miles, respectively for

Route 9 westbound. When 5-SR is used, the average cumulative distance is 14.738 miles. It can be

assumed that the 5-SR distance calculation is the most accurate estimation of actual bus travel distance;

the distance calculated using Google Maps was 14.74 miles. Therefore, the train, pattern, and stop based

distance errors are -0.85%, 0.52%, and -1.7%, respectively. From the AVL/APC data set, pattern distance

is most accurate and was used as the cumulative distance to calculate other related metrics on the

AVL/APC data.

Dwell time is a directly recorded metric included in the archived AVL/APC data. At each stop, the

number of seconds of dwell is recorded, as the time that the door was open. This is not the case with 5-SR

data. However, stop time information can still be gleaned from the data. Within this data set, timestamps

and GPS are not always recorded every 5-seconds; gaps of greater than 5 seconds are seen when the bus

either stops or is moving slowly. Unfortunately, the data does not indicate which of these scenarios

initiates gaps in the data recorded since speeds of zero rarely appear in the data. Therefore, the

assumption must be made that gaps with calculated speeds of 5 mph indicate a stop. One way to estimate

bus stop time is to calculate the change in time between each successive entry to determine the gap time.

If the value is 5 seconds or less, no stopping time is indicated. When the change in time is greater than 5

seconds, that time minus 5 seconds indicates the time spent stopped or in slow motion (gap-stop time).

The average cumulative gap time indicated by the 5-SR data was 29.5 minutes while the BDS data gave

an average dwell time of 20.6 minutes. These two times indicate that the average bus spent almost 9

minutes stopped at locations not associated with passengers boarding or alighting.

0.00

0.50

1.00

1.50

2.00

2.50

3.00∆

in D

ista

nce

(m

i)

5-SR Data

BDS Data


TRIP SPEED ANALYSIS

While the amount of gap-stop time spent at 0 mph remains unclear, a histogram of bus speed could be

created. The speeds of the buses were counted by grouping all reported values into 2 mph bins. The

calculated speed was an average over the time period and did not take into account acceleration or

deceleration. From the grouping of speed data, it was determined that a majority of the buses’ time was

spent moving at less than 10 mph with a quarter of that time traveling less than 1 mph. Figure 5 shows the

breakdown of speeds in 2 mph bins for the average complete trip (n=15) and for all trips between 6:00

a.m. – 10:00 a.m. (n=22). Note that the average trip length was 1 hour 17 minutes.

FIGURE 5 Analysis of time spent moving in speed ranges

This analysis reveals a trend for the buses to be moving less than 5 mph for more than 45% of the

time. In order to identify where these low speeds occur along the route, an analysis of trip speed plotted

against cumulative distance was conducted. More details about bus travel speeds between stops are not

readily available from AVL/APC data as speeds can only be determined between recorded stops. For 5-

SR data, the average distance between data points was about 1/40 mile (132 ft.) rather than the 910 ft.

average stop spacing from the AVL/APC system. This 132 ft. interval was used to define a cumulative

distance and the function needed to create an average.

From data observation, it is clear that in almost all cases there is only one reported speed per bus per

interval When a bus speed was reported for a cumulative distance within each 1/40 mile grouping (e.g.

[7.900 mi –7.925 mi), [7.925 mi–7.950 mi), etc.), the value was added to a table. If a particular trip did

not have a report within the given range, the cell was left blank. When a value was added to the table, a

weight for that value was calculated and then also added to the table. The values of the weights are equal

to the number of seconds that a speed was maintained within a given cumulative distance interval. For

example, if three buses report speeds for the same 1/40 mile segment of 22.5 mph, 9.0 mph, and 1.4 mph

maintained for 4 seconds, 10 seconds, and 64 seconds, respectively, a weight of 4, 10, and 64 is assigned

to each, respectively. The weighted average speed for that segment would be 3.46 mph. Table 2 shows a

sample of how average speed was calculated. S1, S2, etc. are speeds for each bus while W1, W2, etc. are

weights for those speeds.

While only 6 columns of speeds and 6 columns weights are shown in Table 2, 22 bus trips were used

to create Figure 6 which shows the calculated speed versus distance created from both data sets with

major intersection locations noted. The solid black line is the weighted average speed created using 5-SR

data of 22 trips. The dashed grey line is the average speed created by using the AVL/APC data for the 15

complete trips. The mean number of bus speeds recorded per segment was 14.4 with a standard deviation

of 4.6. Figure 6A B, and C show speeds for mile 0 – mile 5.5, mile 5.0 – mile 10.5, and mile 10.0 – end,

respectively.

02468

101214161820

0-1

.99

2-3

.99

4-5

.99

6-7

.99

8-9

.99

10

-11

.99

12

-13

.99

14

-15

.99

16

-17

.99

18

-19

.99

20

-21

.99

22

-23

.99

24

-25

.99

26

-27

.99

28

-29

.99

30

-31

.99

32

-33

.99

34

-35

.99

36

+

Min

s at

Eac

h S

pe

ed

Ran

ge

A: Speed Range (mph)

15%

31%

12%

9%

9%

8%

9%7%

0-0.99 1-4.99 5-9.9910-14.99 15-19.99 20-24.9925-29.99 30+


TABLE 2 Sample of Average Speed Analysis

Distance Weighted

Avg. Speed

Speed

Count

Weight

Count S1 S2 S3 S4 S5 S6 W1 W2 W3 W4 W5 W6

7.900 17.9 6 30 13.5 15.1 18.2 17.3 23.0 20.5 5 5 5 5 5 5

7.925 4.6 6 74 17.0 13.3 7.7 14.9 0.5 10.2 5 5 5 5 49 5

7.950 7.0 4 30 13.3 --- 13.4 1.0 12.1 --- 5 --- 5 15 5 ---

7.975 2.8 5 191 --- 2.1 9.4 14.6 10.1 2.0 --- 81 5 5 5 95

8.000 3.2 4 119 --- 30.8 2.0 10.9 1.7 --- --- 4 45 5 65 ---

8.025 5.3 5 139 2.5 --- 23.5 4.0 28.0 24.8 94 --- 5 30 5 5

8.050 28.9 4 20 28.1 31.5 25.9 --- --- 30.2 5 5 5 --- --- 5

8.075 29.1 2 10 --- --- --- 29.3 28.9 --- --- --- --- 5 5 ---

8.100 20.4 6 31 30.2 12.8 13.0 30.3 27.7 9.9 5 6 5 5 5 5

The speed between stops was also calculated from the AVL/APC data by dividing the distance

between two successive stops by the difference between arrive time at the second stop and the departure

time at the first stop. It should be noted that even though the buses had dwell time where their speed

would be 0 mph, this was not shown in Figure 6. The distance between of locations with AVL/APC data

creates uncertainty in speeds calculated from that data alone. Due to the higher number of reports at

positions between stops locations of AVL/APC, speeds calculated from 5-SR data have a higher

resolution and can be used to examine trip characteristics previously obscured by the lower resolution

AVL/APC data. The new 5-SR data allows a better detection of congestion or delays at intersections. The

analysis also reveals a trend that congestion is prevalent before and after crossing the Ross Island Bridge

but traffic still runs smoothly on the bridge itself.

NW

Eas

tman

P

kwy

SE 1

74

th

SE 1

48

th

SE 1

36

th

SE 2

70

0

Blo

ck

SE 1

82

nd

SE 1

64

th

0

5

10

15

20

25

30

35

40

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 5.50

Spe

ed

(m

ph

)

a) Cumulative Distance Start - 5.5 (miles)

Speed (from 5-SR data) Speed (from BDS data)

SE 1

36

th

SE 1

22

nd

SE 1

12

th

SE 9

9th

SE 9

2n

d

SE 8

2n

d

SE 5

2n

d

SE 3

9th

SE 7

1st

0

5

10

15

20

25

30

35

40

5 6 7 8 9 10

Spe

ed

(m

ph

)

b) Cumulative Distance 5-10.5 (miles)


FIGURE 6 Average speed versus cumulative westbound trip distance, n=22

The two data series in Figure 6 do not contain all the same information at different resolutions. Since

distance is used for the x-axis, only non-zero speeds permit the function to continue, the AVL/APC data

lacks stop time consideration while the 5-SR plot does not since slow speeds associated with long gap

time are taken into account. For example, at distance 7.900 in Table 2, all 6 speeds are similar and have

equal weight because each speed was recorded for a 5-second interval while at 8.000 the largest speed has

a weight of 0.8 while the slowest speed has a weight of 13 indicating speed duration of 4-seconds and 65

seconds, respectively. Distance 8.000 marks the crossing of SE 82nd on Powell; while all the buses

stopped at this location, zero speeds tend not to appear while using 5-SR data. However, it can be

assumed that distances where the average speed falls below 5 mph likely included stopped buses even if

the exact stop time is unknown.

FIGURE 7 Average speed versus cumulative distance trend line, n=22

As shown in Figure 7, the analysis of average speed allowed for two linear trend lines to be created:

one for the stop-level AVL/APC data and one for 5-SR. These trend lines indicate that the speed of buses

decreases as they move westward during the morning commute. An average speed estimate can be

calculated by integrating the linear regression lines over the distance and dividing by total distance

traveled. This average speed divided by total distance gives an average trip time estimate. The AVL/APC

and 5-SR data sets resulted in calculated average trip times of 1 hour 4.7 minutes and 1 hour 9.5 minutes,

respectively. Since it is known that the mean trip time was 1 hour 17 minutes, an error of >10% is

associated with each estimate.

SE 3

9th

SW A

rth

ur&

1st

Beg

in

Ro

ss Is

lan

d

End

Ro

ss Is

lan

d

SE 2

1st

SW 6

th &

Mill

SW 6

th &

Ald

er

End

of

Ro

ute

0

5

10

15

20

25

30

35

40

10 11 12 13 14 15

Spe

ed

(m

ph

)

c) Cumulative Distance 10-End (miles)

NW

Eas

tman

Pkw

y

SE 1

74

th

SE 1

48

th

SE 1

36

th

SE 1

22

nd

SE 1

12

th

SE 9

9th

SE 9

2n

d

SE 8

2n

d

SE 5

2n

d

SE 2

70

0B

lock

SE 1

82

nd

SE 3

9th

SW A

rth

ur

& 1

st

Beg

in R

oss

Isla

nd

End

Ro

ss Is

lan

d

SE 2

1st

SE 1

64

th

SE 7

1st

SW 6

th &

Mill

SW 6

th &

Ald

er

End

of

Ro

ute

y = -0.8303x + 23.293

y = -0.9395x + 22.917

0

5

10

15

20

25

30

35

40

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Spe

ed

(m

ph

)

Cumulative Distance (miles)


Despite this error, the trip speed versus distance graph allows for the locations of unscheduled stops

to be observed. For example, the quarter mile running up to mile SE 82nd (mile 8.0) runs slow preceding

the stop location. This is likely caused by buses waiting in queues before crossing SE 82nd to reach the

stop on its far side.

TRIP TIME ANALYSIS

Figure 8 was created using similar methodology employed in the creation of Figure 6. The x-axis is the

actual time and the y-axis represents an average speed created from all bus trips operating over all

positions along the route at the same time. The grey line is an average speed at 5 second intervals. The

black line is 1-minute moving average speed with ±30 seconds of accuracy. The white line is a

polynomic trend line that highlights the overall shape of the plot. A dip in the speeds between 7:00 a.m.

and 9:00 a.m. coincides with the morning congestion and represents a decrease in average speed of about

9 mph.

FIGURE 8 Average bus speed versus time, 2≤n≤6

A trip time analysis using 5-SR data has some limitations created by the lack of information regarding

passenger movement (i.e. passenger boarding and alighting), specified dwell time, use of a lift, traffic

signal indications, activation of transit signal priority, etc. Without more independent information, a trip

time model using 5-SR data will be of limited utility. The observed decrease in speed serves to confirm

the effects of morning congestion on Route 9 buses.

INTERSECTION LEVEL ANALYSIS

In the previous section, it was noted that around SE 82nd and Powell Blvd., buses have lower speeds; SE

82nd is also Oregon route 213 and has higher traffic signal coordination priority than Powell Blvd. The

area from SE 84th – SE 80th was examined in detail and is shown in Figure 9, average bus speed is

decreasing as they approach SE 82nd. Bus speed increases after the bus has passed its scheduled stop on

the far side of the intersection (most buses stop at 82nd). The speed of the buses is steady around 27 mph

until the buses pass SE 80th Ave. and approaches the bus stop at SE 79th Ave. This example shows that it

is now possible to zoom in into specific intersections and detect areas with significant queuing.

0

5

10

15

20

25

30

35

40

Spe

ed

(m

ph

)

5-Second average 1-Minute Average


FIGURE 9 Speed Analyses of SE Powell & 82nd

CONCLUSIONS

The results of this study suggest that the new generation of higher resolution bus trajectory data can be

successfully employed to identify congestion along urban arterials.

The analysis conducted shows that 5-SR data can be used to observe metrics about operating speed in

more detail than could previously be seen using stop-level AVL/APC data. This study found that the

average travel speed decreased as buses moved eastward on Route 9 with an overall average of 17.1 mph.

Additionally, while it is to be expected that slow average bus speeds should occur at scheduled bus stops,

slow speeds were reported on approaches to many bus stops. This indicates congestion before buses were

able to reach their stop destinations, especially for bus stops shortly following signalized intersections

such as SE 82nd Ave. and SE 39th Ave. An analysis of average speed created using 5-SR data at each

intersection can indicate where buses are stopping and highlight whether those locations are intended to

be slow moving or a stop.

Dwell time accounted for 27% of the average trip time of 1 hour 17 minutes. Gap-stop time

accounted for 38% of this average trip. The speed breakdown shows that 46% of the time was spent

moving <5 mph; therefore, it can be concluded that 27%-38% of the time was spent stopped. However,

the exact stop time remains uncertain.

The next step is to validate this model by examining specific intersections for all buses for a complete

day. This will help to determine more specifically where problems are occurring and allow for solutions

to be presented. This analysis calls for a recommendation for a change in TriMet’s 5-SR data. It is

recommended that reports should be made every 5 seconds regardless of bus motion. This will allow for

accurate stop times and the locations of these stops to be directly analyzed. Currently, only assumption of

slow speed can be used and actual stopping time is uncertain. In addition, it should be noted that without a

few additional pieces of information, 5-SR data is not accessible on its own. It requires that BDS data be

used to compare and extract the data. This could be resolved by including additional fields in the data

about train and trip number. Wheel sensor movement data could be another complementary dataset to

overcome the limitations in this paper.

ACKNOWLEDGEMENTS

We would like to express gratitude to Steve Callas of TriMet for graciously providing both data sets used

for analysis and to the peer reviewers for their helpful and encouraging comments.

SE 84th

Beg

in S

E 8

2n

d

End

SE

82

nd

Bu

s St

op

82

nd

SE 80th

0

5

10

15

20

25

30

7.86007.88007.90007.92007.94007.96007.98008.00008.02008.04008.06008.08008.10008.1200

Ave

rage

Sp

eed

(m

ph

)


REFERENCES

1. Levinson, H. S. 1983. “Analyzing Transit Travel Time Performance.” Transportation Research

Record, no. 915.

2. Cambridge Systematics. 1999. Multimodal Transportation: Development of a Performance-Based

Planning Process. NCHRP Project No. 8–32(2).

3. Hall, R., and N. Vyas. Buses as a Traffic Probe. Transportation Research Record No. 1731, 2000, pp.

96–103.

4. Chakroborty, P., and S. Kikuchi. Using Bus Travel Time Data to Estimate Travel Times on Urban

Corridors. Transportation Research Record No. 1970, 2004, pp. 18–25.

5. Bertini, R. L. and S. Tantiyanugulchai. Transit Buses as Traffic Probes: Empirical Evaluation Using

Geo-location Data. Transportation Research Record No. 1870, 2004, pp. 35–45.

6. Berkow. M., Wolfe, M., Monsere, and C.M., Bertini, R. L. and S. Tantiyanugulchai. Using Signal

System Data and Buses as Probe Vehicles to Define the Congested Regime on Arterials, In

Transportation Research Record: Journal of the Transportation Research Board, No. 1870, TRB,

National Research Council, Washington, D.C., 2004, pp. 35–45.

7. Feng, Wei. 2014. “Analyses of Bus Travel Time Reliability and Transit Signal Priority at the Stop-

To-Stop Segment Level.” Dissertations and Theses, June.

8. Strathman, J. G., K. J. Dueker, T. Kimpel, R. L. Gerhart, K. Turner, P. Taylor, S. Callas, and D.

Griffin. 2000. “Service Reliability Impacts of Computer-Aided Dispatching and Automatic Vehicle

Location Technology: A TriMet Case Study.” Transportation Quarterly 54 (3): 85–102.

9. Strathman, J. G., T. J. Kimpel, K. J. Dueker, R. L. Gerhart, and S. Callas. 2002. “Evaluation of

Transit Operations: Data Applications of Tri-Met’s Automated Bus Dispatching System.”

Transportation 29: 321–45.

10. Strathman, James, Kenneth Dueker, Thomas Kimpel, Rick Gerhart, Ken Turner, Pete Taylor, Steve

Callas, David Griffin, and Janet Hopper. 1999. “Automated Bus Dispatching, Operations Control, and

Service Reliability: Baseline Analysis.” Transportation Research Record 1666 (1): 28–36.

11. Data Resources Derived from BDS Card Data. 2003. Stop Event Data Dictionary.

12. Bertini, R. L., and A. M. El-Geneidy. 2004. “Modeling Transit Trip Time Using Archived Bus

Dispatch System Data.” Journal of Transportation Engineering 130 (1): 56–67.

13. Bertini, R. L., and A. El-Geneidy. Generating Transit Performance Measures with Archived Data.

Transportation Research Record No. 1841, 2003, pp. 109–119.

14. Bertini, R. L., and Li, H. 2011. "Comparison of Algorithms for Systematic Tracking of Patterns of

Traffic Congestion on Freeways in Portland, OR." Transportation Research Board.

Documents

Exploring Applications of Second Generation Archived